Products > Thermal Imaging

FLIR P640 thermal Image distorted vertical lines on screen

<< < (5/5)

Fraser:
A quick look at the Patent makes me think I was on the right track with your issue being an imaging processing problem. It does not appear that the Column amplifier gains are involved in the image noise processing stages. The “RAW” image information coming out of the Microbolometer is ‘warts and all’ but the various biases applied to the microbolometer are set to provide the optimum performance in the desired role or scenario. Different temperature ranges may be configured with appropriate biases applied. The “RAW” image undergoes ‘treatment’ by the video processing stages but the Radiometric measurement data is preserved. The Row and Column filtering is intended to greatly improve the image by effectively removing the horizontal and vertical line ‘noise’ as detailed in the a patent I have referenced. This noise reduction is ‘intelligent’ in that that it ‘studies’ the image data resented to it in the form of pixel values and makes appropriate adjustments based on adjacent pixel values. It is effectively ‘hunting’ for classic vertical and horizontal lines associated with rows an columns rather than straight edges in the thermal scene. I will need to study the Patent and think about this more. The old brain is tired and needs a rest.

Bill W:

--- Quote from: Fraser on November 28, 2021, 11:34:23 pm --- The Row and Column filtering is intended to greatly improve the image by effectively removing the horizontal and vertical line ‘noise’ as detailed in the a patent I have referenced. This noise reduction is ‘intelligent’ in that that it ‘studies’ the image data resented to it in the form of pixel values and makes appropriate adjustments based on adjacent pixel values. It is effectively ‘hunting’ for classic vertical and horizontal lines associated with rows an columns rather than straight edges in the thermal scene. I will need to study the Patent and think about this more. The old brain is tired and needs a rest.

--- End quote ---

So if what appears to the image analysis as a 'hot column' is a part of the real image, the software will try to 'fix' the whole column and cause problems to real bland areas where the 'hot column' is not ?
That sounds very close to the OP observations.

May be something of more use in shutterless applications as it is column patterns of divergence that tend to show up first, as the amplifiers slowly drift away from the last calibration.

radioorizont:
Did a bit more work, In the calibration file i changed image settings of the digital filter from column threshold 60 to 15, left the thresholdPix at 120 default.
I did this for the -40 to 150 range (3 lines of code needing data changed)
The images are now much better with no vertical lines.
The only small thing is at high sensitivity with low temperature spans you can see some vertical noise
If i go up on threshold the vertical noise will be hidden but then the lines artefact gets introduced, it's a compromise.
I am happy with how the camera works now, i might look at other tweaks to further improve img quality.
Photos attached with outcome and with a noisy image when pointing at wall with slight vertical shape on left side (no temperature difference so gain at highest)
Very happy with how it looks considering the mod, left the 100 to 500 unchanged, might mod that too later, easy to revert back any time.

Fraser:
It would be worth asking Ultrapurple what his cameras column filter settings are set to. It might be slightly different to your camera but will give you an idea of the ‘nominal’ values.

When buying a used thermal camera of this sophistication, it can be the case that it has been fine tuned in the configuration for a particular application and may not then produce good imagery in another application until a more general configuration is applied.

At the end of the day, as long as you are happy with the cameras performance, that is all that matters  :-+

Fraser

Navigation

[0] Message Index

[*] Previous page

There was an error while thanking
Thanking...
Go to full version